Internal Link Structure: Models That Actually Work

Internal link structure is the shape of your site's link graph: which pages link to which, and how many clicks separate any page from the homepage. The structure determines how link equity spreads, how fast crawlers find new content, and whether users can navigate by topic instead of by search box. There is no single correct model, but there are four named patterns worth knowing, one hard constraint (click depth), and a sensible default for each site type.

The hard constraint first: click depth

Click depth is the minimum number of clicks from your homepage to a given page, following links. It matters for three concrete reasons:

1. Equity decays with depth. Under any PageRank-like model, each hop splits and dampens the authority flowing along it. A page five hops out receives a fraction of what a page two hops out receives from the same source.
2. Crawl priority follows depth. Search engines crawl shallow pages more often. Deep pages get discovered later and refreshed less frequently, a real cost when you update content.
3. Users give up. Every additional required click loses a share of visitors.

The working rule: important pages within 3 clicks of the homepage, everything indexed within 4–5. This isn't a Google-published threshold; it's arithmetic. A homepage linking to 100 pages, each linking to 100 more, reaches a million pages in three hops. If your 5,000-page site has content at depth 7, the structure, not the size, put it there.

Depth is also the structural number a crawl measures most reliably. LinkAgent computes click depth for every page during its crawl and uses it directly in scoring: links proposed from shallow, strong pages are worth more, and deep, link-starved targets get priority. Depth problems are structure problems, and they show up before any anchor or content problem does.

The four models

Flat structure

Every page sits one or two clicks from the homepage: think a 30-page consultancy site where the nav reaches everything.

• Strengths: minimal depth, every page gets equity, nothing hides.
• Weakness: doesn't scale. Past a few dozen pages, "flat" means a 200-link homepage, which splits equity into confetti and gives users a phone book. Flatness also encodes no topical relationships: every page is equally related to every other, which is to say, not related at all.
• Use when: the site is genuinely small (< ~50 pages) and will stay that way.

Hub-and-spoke (topic clusters)

A hub page covers a topic broadly and links to each spoke (a page covering one subtopic in depth); every spoke links back to the hub and to sibling spokes where relevant. A hub on "espresso" links to grind size, tamping, pressure profiling, machine maintenance; each of those links back and across.

• Strengths: encodes topical relationships explicitly; concentrates equity on the hub (usually the page targeting the hardest keyword); gives every new spoke an immediate linked home. This is the same architecture the topic-cluster and modern siloing crowd converged on; the comparison is in SEO silos without the straitjacket.
• Weakness: requires real hub pages: a thin category listing doesn't earn the position. And clusters need cross-links between them or the site fragments into islands.
• Use when: content maps to topics with a head term and long-tail subtopics. Most content-led sites.

Pyramid (hierarchy)

Homepage → sections → subsections → detail pages, with links flowing mostly parent↔child. The classic ecommerce shape: home → category → subcategory → product.

• Strengths: scales indefinitely; depth stays logarithmic in page count; matches how users think about inventories ("men's → running → trail").
• Weakness: left to itself, a pure pyramid has almost no lateral links, so equity pools in categories while detail pages starve, and topically related leaves in different branches never reference each other.
• Use when: large inventories with natural taxonomy; then deliberately add lateral links (related products, "pairs with", buying guides linking into product pages).

Mesh

Every page links freely to every related page, no enforced hierarchy; wikis are the canonical example.

• Strengths: maximal discoverability; equity spreads to wherever relevance exists; extremely robust (no single hub failure orphans a branch).
• Weakness: with no hierarchy, nothing concentrates. If every page is equally linked, no page is emphasized, and the site ranks broadly-but-shallowly. Meshes also degrade without discipline: link everything to everything and anchors go generic.
• Use when: reference content where users arrive mid-graph from search (docs, glossaries, wikis), ideally layered over a light hierarchy rather than replacing one.

Picking a model by site type

Blog / content site → hub-and-spoke. Group posts into clusters with a genuine pillar page per cluster. The chronic failure mode of blogs is the reverse-chronological archive as the only structure, which buries older posts at ever-increasing depth. Hubs fix that: a post's depth becomes home → hub → post, permanently 2, regardless of publish date.

Ecommerce → pyramid plus deliberate lateral links. The taxonomy gives you the pyramid for free. Your work is (a) keeping depth down when subcategories nest (use subcategory links and curated "top products" links from higher levels), and (b) adding the lateral layer: buying guides that link to products, product pages cross-linking alternatives, category descriptions that are actual prose with actual links rather than three boilerplate sentences.

SaaS / marketing site → small pyramid for the product pages, hub-and-spoke for content, and heavy bridges between them. The most common SaaS structure failure is two disconnected graphs: a polished 20-page marketing pyramid and a 300-post blog cluster that links to the product exactly once, from the nav. The bridges (contextual links from relevant posts to feature and solution pages, with descriptive anchors) are where blog equity becomes commercial rankings.

Docs → shallow hierarchy plus mesh. A sidebar hierarchy for orientation, dense in-prose cross-references for actual navigation, because doc users land mid-graph from search and move laterally between related concepts.

Auditing the structure you actually have

Every site has a link structure; few have the one their information architecture diagram claims. The audit:

1. Crawl and compute depth per page. Sort descending. Anything important at depth ≥ 4 is a finding.
2. Look at the contextual graph only. Strip navigation, footer, and sidebar links from the analysis (LinkAgent does this by construction: its graph contains only in-text links). Sites that look healthy with nav links included often reveal dozens of contextual orphans without them.
3. Map inbound in-text link counts against page priority. The classic finding: money pages near zero, a random 2019 post at forty.
4. Check cluster integrity. Do spokes link back to hubs? Do siblings interlink? Do clusters bridge to commercial pages?
5. Fix with links before restructuring URLs. Most "site restructure" projects should be link projects. Adding hub pages and cross-links changes the graph without migrations, redirects, or risk. URL restructuring (a different lever, often conflated with this one) is covered in site structure for SEO.

Doing step 3–5 by hand means scoring hundreds of candidate page pairs for relevance. This is the part worth automating: an internal linking tool that builds the contextual graph, clusters pages by topic (TF-IDF similarity), and proposes links scored on source depth and target link need (with anchors taken verbatim from existing sentences and every proposal passing through an approve/reject queue) turns a quarter-long restructure into an afternoon of review.

Keeping the structure alive

Structures decay one published post at a time. Each new page needs a cluster assignment, links in from the hub and 2–3 relevant siblings, and links out to the pages it discusses, at publish, not at next year's audit. Scheduled re-crawls (LinkAgent supports monthly, weekly, or daily) catch the drift automatically and propose links for new pages as they appear.

Structure is the skeleton of the broader discipline: anchors, equity flow, and maintenance hang off it. For the end-to-end system, start with the complete internal linking guide; pick your model from this post, and let depth be the number you manage.